CN110794413B - Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels - Google Patents
Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels Download PDFInfo
- Publication number
- CN110794413B CN110794413B CN201911105764.6A CN201911105764A CN110794413B CN 110794413 B CN110794413 B CN 110794413B CN 201911105764 A CN201911105764 A CN 201911105764A CN 110794413 B CN110794413 B CN 110794413B
- Authority
- CN
- China
- Prior art keywords
- linear
- point cloud
- power line
- voxel
- voxels
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
- G01S17/06—Systems determining position data of a target
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
Abstract
The invention relates to a power line detection method and system, belongs to the technical field of power detection, and particularly relates to a method and system for detecting a laser radar point cloud data power line by linear element segmentation. The method comprises the steps of collecting three-dimensional point cloud data of a surface power line by using an airborne laser radar system and a POS system, filtering the point cloud data to extract ground object points, converting the point cloud data into voxels based on the idea of linear voxel segmentation, and realizing the extraction of power line segments by detecting the linearity of the voxels and the positioning of a tower, thereby achieving the aim of accurately extracting the power line point cloud from the point cloud data.
Description
[ technical field ] A method for producing a semiconductor device
The invention relates to a power line detection method and system, belongs to the technical field of power detection, and particularly relates to a method and system for detecting a laser radar point cloud data power line by linear element segmentation.
[ background of the invention ]
As power demand has increased dramatically in today's life and industry, modeling and hazard monitoring of power line networks has become critical. Power lines are typically several tens of kilometers long and are built in harsh terrain environments, which can be a technical challenge to detect. At present, power line network operators generally use helicopters carrying infrared cameras to monitor most of the transmission lines; for distribution lines and corridors thereof with the voltage level of 110KV or lower, technicians walk along the power lines to perform manual inspection, but the method is expensive and long in time consumption, and detection results often have errors.
As a novel active earth observation System, an airborne laser radar (LiDAR) System is taken as advanced equipment such as a highly integrated Global Positioning System (GPS), an Inertial Navigation System (INS) And a laser scanning distance meter, the high-precision And high-density three-dimensional point cloud on the surface of a ground object at sea can be obtained in a short time, And a laser beam emitted by the LiDAR System can form multiple echoes. The airborne LiDAR system is utilized to carry out electric power line patrol, so that the method is efficient and quick, and the defects that the electric power line patrol workload is large, the danger is high, the efficiency is low and the space positioning precision of the helicopter electric power line patrol is low in the traditional engineering measurement are overcome. The location information about each conductor, the geometry information of the tower, the topographical profile, and the location and shape information of all terrain (including trees, buildings, and other terrain) scanned by the LiDAR system may be used to detect conductor damage, structural changes to the tower, and to find trees that are potentially dangerous to the power line, thereby better planning maintenance work on the power line.
In recent years, with the continuous development and popularization of LiDAR technology, more and more power departments or units adopt the LiDAR technology to realize power line patrol business. The existing method for extracting the power line has large calculated amount in a large power line three-dimensional scene, does not have universal applicability, simultaneously fails to fully consider the context relationship between a tower and the power line, has the problem that the automatic extraction precision cannot meet the high engineering application requirement, and needs more manual intervention. Therefore, the method for accurately detecting the power line point cloud directly from the three-dimensional LiDAR point cloud data has important practical significance for power line patrol work.
[ summary of the invention ]
In view of the above, in order to overcome the defects of the prior art, the invention provides a method and a system for detecting a power line of laser radar point cloud data by linear voxel segmentation, which can quickly and accurately extract the power line point cloud data, and aims to solve the problems of complex processing flow, long time consumption and the like in the process of automatically classifying the point cloud and then manually editing the power line point cloud in the prior art.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a method for detecting a laser radar point cloud data power line by linear voxel segmentation comprises the following steps:
a ground object point cloud extraction step, wherein ground point cloud is separated from three-dimensional point cloud data acquired by a flight platform, so that ground object point cloud data are obtained;
a linear voxel segmentation step, wherein a linear segment map model is constructed based on linear voxel segmentation;
a candidate line segment extraction step, namely extracting a power line candidate linear segment by adopting a Markov random field model;
and a power line point cloud detection step, namely positioning the power line linear section based on the context relationship between the tower and the power line, and extracting the power line point cloud through the distance threshold value between the power line point cloud and the power line section.
Preferably, in the above method for detecting a power line of linear voxel partitioned lidar point cloud data, in the step of extracting the point cloud of the ground object, the obtaining of the point cloud of the ground object includes:
a triangular mesh construction sub-step, namely constructing a mesh index for the point cloud data, and taking the lowest point of each block in the mesh as an initial ground seed point; constructing an initial triangulation network TIN based on the initial ground points;
and in the point cloud traversing substep, traversing the residual point cloud, and adding the residual point cloud which meets preset conditions relative to the ground height and the terrain angle into the triangular network.
Preferably, in the method for detecting a laser radar point cloud data power line by linear voxel segmentation, the step of waiting for linear voxel segmentation includes:
a voxel processing substep, which is to perform voxel processing on the ground object point cloud data by utilizing a multi-branch tree structure;
a principal component analysis substep, namely constructing a covariance matrix for the point cloud in each voxel, calculating a characteristic value through a principal component analysis method, and identifying a linear structure point in the voxel based on the characteristic value;
a linear structure evaluation sub-step of screening the voxels having a linear structure based on the proportion of the linear structure points within the voxels.
A map model generation sub-step of converting points within the linear voxels into linear segments and generating a map model in combination with the linear segments and the connections therebetween.
Preferably, in the above method for detecting power lines of point cloud data of laser radar for linear body segmentation, in the step of extracting candidate line segments, candidate linear segments of power lines are extracted based on a markov random field model, and the potential energy at the vertex is calculated by using the height Δ H from the ground of the linear segment, the line slope κ of the linear segment, and the parallelism σ of the inner and outer cylindrical points μ of the linear segment, and an energy function E is constructed, and the candidate linear segments of power lines are searched by minimizing the energy function, where the potential energy function U and the energy function E are based on the following formulas:
where U is the potential energy of the apex. Alpha is alpha1,α2,α3,α4Are weight coefficients. The Q function calculates the likelihood distribution of the vertex, the V function calculates the spatial correlation of the vertex in the neighborhood range, if the labels of the adjacent vertexes are different, the V is endowed with a penalty coefficient, and the f is the optimized label.
Preferably, in the method for detecting a power line of point cloud data of a laser radar based on linear voxel segmentation, the step of detecting the point cloud of the power line includes:
a power line model construction sub-step, namely positioning the position of a tower based on the context of a linear section; constructing a connecting edge set of a tower, namely a power line model;
and a power line extraction sub-step, namely classifying the ground object point cloud data, and extracting the power line point cloud by setting a distance threshold between the point cloud and the power linear section.
A laser radar point cloud data power line detection system for linear voxel segmentation comprises:
the ground object point cloud extraction module is used for separating ground point cloud from the three-dimensional point cloud data acquired by the flight platform so as to obtain ground object point cloud data;
a linear voxel segmentation module for constructing a linear segment map model based on linear voxel segmentation;
the candidate line segment extraction module is used for extracting a power line candidate linear segment by adopting a Markov random field model;
and the power line point cloud detection module is used for positioning the power line linear section based on the context relationship between the tower and the power line and extracting the power line point cloud through the distance threshold value between the power line point cloud and the power line section.
Preferably, in the above system for detecting power lines of linear voxel division laser radar point cloud data, the ground object point cloud extraction module further includes:
the triangular net constructing unit is used for constructing a grid index for the point cloud data, and taking the lowest point of each block in the grid as an initial ground seed point; constructing an initial triangulation network TIN based on the initial ground points;
and the point cloud traversing unit is used for traversing the residual point cloud and adding the residual point cloud which meets the preset conditions with respect to the ground height and the terrain angle into the triangular network.
Preferably, in the above system for detecting a laser radar point cloud data power line for linear voxel segmentation, the candidate linear voxel segmentation module includes:
the voxel processing unit is used for carrying out voxel processing on the ground object point cloud data by utilizing a multi-branch tree structure;
the principal component analysis unit is used for constructing a covariance matrix for the point cloud in each voxel, calculating a characteristic value through a principal component analysis method, and identifying a linear structure point in each voxel based on the characteristic value;
and a linear structure evaluation unit for screening the voxels having a linear structure based on the proportion of the linear structure points within the voxels.
And a graph model generation unit for converting the points in the linear voxels into linear segments and generating a graph model by combining the linear segments and the connections therebetween.
Preferably, in the above system for detecting power lines of point cloud data of laser radar for linear element segmentation, the candidate line segment extraction module extracts a candidate linear segment of a power line based on a markov random field model, calculates a potential energy U at a vertex by using a height Δ H from a ground of the linear segment, a line slope κ of the linear segment, and a parallelism σ of inner and outer cylindrical points μ of the linear segment, constructs an energy function E, and searches for the candidate linear segment of the power line by minimizing the energy function, where the potential energy function U and the energy function E are based on the following formulas:
where U is the potential energy of the apex. Alpha is alpha1,α2,α3,α4Are weight coefficients. The Q function calculates the likelihood distribution of the vertex, the V function calculates the spatial correlation of the vertex in the neighborhood range, if the labels of the adjacent vertexes are different, the V is endowed with a penalty coefficient, and the f is the optimized label.
Preferably, the above laser radar point cloud data power line detection system for linear element segmentation includes:
the power line model building unit is used for positioning the position of the tower based on the context of the linear section; constructing a connecting edge set of a tower, namely a power line model;
and the power line extraction unit is used for classifying the ground object point cloud data and extracting the power line point cloud by setting a distance threshold between the point cloud and the power linear section.
Compared with the prior art, the invention has the following advantages:
1. the point cloud detection method has simple processing flow and obviously improved calculation efficiency, and is suitable for large power line three-dimensional scenes;
2. the context relation between the power line and the tower is considered in the detection of the power line point cloud, the power line segment is further brushed and selected through the accurate positioning of the tower, then the detection of the power line point cloud is carried out, and the detection precision of the power line point cloud is improved.
[ description of the drawings ]
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of the power line point cloud extraction method of the present invention.
Fig. 2 is a schematic diagram of a point cloud voxelization process based on an octree structure.
FIG. 3 is a schematic diagram of the ratio structure of inner and outer cylindrical points.
[ detailed description ] embodiments
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the method for detecting a power line of point cloud data of a lidar for linear voxel segmentation in the embodiment includes:
the method comprises the steps of (1) point cloud collection, wherein airborne laser radar data are collected and three-dimensional point cloud data are generated by utilizing an airborne laser radar and a POS system which are configured and installed on a flight platform;
step (2), a ground feature point extraction step, wherein ground point clouds are separated based on an iterative triangulation encryption filtering algorithm to obtain a ground feature point cloud set;
step (3), a linear segmentation step, namely performing voxelization processing on ground object point cloud data by using an octree structure, and performing linear detection on the basis of the voxelization structure to construct a linear segment map model;
step (4), a power line candidate linear segment extraction step, which is to extract a power line candidate linear segment based on a Markov random field model by taking the ground height, the line slope, the ratio of inner and outer cylindrical points and the line parallelism as data items;
step (5), a tower positioning step, namely positioning a tower based on the context relationship between the tower and the power line, and constructing a connection edge set of the tower;
and (6) extracting power line point clouds, namely automatically extracting the power line point clouds through a set cylindrical search box based on the power line model extracted in the step (5).
In the step 2, the specific steps of extracting the ground object point cloud based on the iterative triangulation encryption filtering are as follows:
step 2.1, the laser radar point cloud data obtained in the step 1 is used as input, and the point cloud data is processed in the step 2.2-2.5;
step 2.2, constructing a grid index for the point cloud data, and taking the size of the maximum building in the scene of the measuring area according to the size of the grid;
step 2.3, aiming at each block in the grid, taking the lowest point of each block as an initial ground seed point, and constructing an initial TIN (triangulated irregular network), namely an approximate surface model, based on the initial ground point;
step 2.4, iteratively judging whether the remaining point cloud meets the condition of adding the remaining point cloud into the terrain TIN, wherein the judgment criterion comprises the following steps:
1) relative ground height is the vertical distance d from the point cloud p to be classified to the terrain triangulation network where the point cloud p is located;
2) terrain angle: the maximum value theta of the included angles between the point cloud p to be classified and the three vertexes of the terrain triangulation network where the point cloud p is located;
if the laser foot point to be judged satisfies d is less than or equal to TdAnd theta is less than or equal to TθIf the point is determined to be a ground point, adding the point to a TIN model, wherein TdAs a relative ground height threshold, TθIs the terrain angle threshold.
And 2.5, in a certain iteration, if no ground point meeting the judgment criterion exists, the surface filtering is finished, and the rest non-ground point cloud is the ground object point.
In step 3, the specific steps of the linear voxel segmentation process are as follows:
step 3.1, carrying out voxelization processing on the ground object point cloud data by using an octree structure, wherein the specific operation comprises the following steps: 1) counting the maximum and minimum values of the point cloud data on the X, Y and Z axes to obtain a three-dimensional bounding box of the point set; 2) the entire three-dimensional bounding box is considered as an initial three-dimensional grid node, divided into regular three-dimensional grids, and points within each cube are defined as voxels. And if more than three points are contained in the grid node, the grid node is defined as full, otherwise, the grid node is defined as empty. And deleting the octree nodes which are empty, and further subdividing the non-empty nodes until a termination standard is reached:
R≤Rvoxel||pnum≤minp (1)
wherein R is the voxel volume and is expressed by the side length of a cube, RvoxelIs the set voxel resolution. p is a radical ofnumIndicating the voxel size, i.e. the number of points in the voxel, minpRepresenting the minimum voxel size and is related to the point cloud density.
As shown in fig. 2, in the present embodiment, the ground object point cloud data is voxelized by using an octree structure.
Step 3.2, the linear likelihood of each voxel is evaluated using Principal Component Analysis (PCA). Firstly, a covariance matrix is constructed for the point cloud in each voxel, and eigenvalues (lambda 1, lambda 2, lambda 3) are calculated through PCA. The point cloud can be divided into linear structure points, surface structure points and scattering structure points based on the spatial distribution of the point cloud, wherein the linear structure points are in linear characteristics in space, and characteristic values obtained through PCA have the conditions that lambda 1 is greater than lambda 2 is greater than lambda 3; points on the surface have planar characteristics in space, and the difference between the characteristic values lambda 1 and lambda 2 is small; scattering points are distributed and scattered in space, so that the values of lambda 1, lambda 2 and lambda 3 are close; can be obtained by using
To identify a linear structure point cloud within the voxel. The determination of the linear structure voxel can be determined by calculating the proportion of the linear structure points in each voxel, if the point clouds in the voxel are all linear structure points, the voxel is a linear voxel, if the non-linear structure points exist in the voxel, the non-linear structure points are compared with the adjacent voxel, and if the adjacent voxel is a linear voxel and has a similar direction with the linear voxel, a linear voxel label is given to the adjacent voxel.
Step 3.3, a random sample consensus (RANSAC) based 3D line fitting converts points within the linear voxels to linear segments and combines the linear segments (i.e., vertex V) and the connection between them (i.e., edge E) to generate a graph model G ═ V, E.
The point cloud voxelization process based on the octree structure is shown in fig. 2.
In step 4, the specific steps of the power line candidate linear segment extraction based on the markov random field model are as follows:
step 4.1, calculating the height (Delta H) and the maximum value H of the linear sectionmaxThe power line is overhead, while the bottom of the tower is in contact with the ground.
Wherein H represents the elevation of the linear section, HGIs the ground level.
And 4.2, calculating the slope k of the line, wherein the power line is horizontal, and the rest parts of the tower except the horizontal cross arm are vertical.
Where N represents the number of point clouds in the linear voxel, κxi、κyiRespectively represents the fitted curve slope of the ith point cloud in the linear voxel in XZ and YZ projection.
And 4.3, calculating the point ratio (mu) of the inner cylinder and the outer cylinder, and creating the inner cylinder and the outer cylinder by taking the line segment as a circular axis, wherein in the embodiment, the inner radius is preferably 0.15m, the outer radius is preferably 1m, the radii are only examples, and any other values suitable for creating the inner cylinder and the outer cylinder can be used in the embodiment. And calculating the ratio of the number of point clouds in the inner and outer circular columns. There is almost no point cloud between the inner and outer columns of the power line.
Where M represents the number of point clouds in the inner cylinder and M represents the number of point clouds in the outer cylinder.
The inner and outer cylindrical point ratio structure is shown in fig. 3.
And 4.4, calculating the parallelism degree sigma of the lines, wherein a single power line is generally parallel to other power lines, but the towers are randomly distributed. Suppose the direction vector of the linear segment p is ξpThe search window is set to 20m in size and the number of linear segments contained in the window isN, the number of linear segments having a direction coincident with that of linear segment p is m (let the direction vector of linear segment q be ξ)qIf 1- | ξq·ξp|<0.1, the linear segments q and p are considered to have a uniform direction), the line parallelism σ of the linear segment p ispCan be expressed as
And 4.5, describing the unique characteristics and the context attributes of the linear segment graph model constructed in the step 3 by adopting a Markov random field, and constructing an energy function by utilizing three data items (the height from the ground, the slope of a line, the ratio of inner and outer cylindrical points) and one context item (the parallelism of the line). In a graph model G ═ (V, E) composed of vertices V (linear segments) and edges E (connected edges of linear segments), only locally connected edges remain in the graph, i.e. if the distance between a pair of vertices connecting the edges is greater than a specified threshold (T)d3m), the edge is discarded. Each energy term has a different energy function according to the state of the current vertex or the adjacent vertex (i.e., power line or non-power line), and the potential energy of the vertex can be calculated by equation (6). The power line candidate linear segment is searched through minimization of the energy function, and the linear segment with the highest energy is switched iteratively until the energy converges. Let the energy function form be as shown in equation (7).
Where U is the potential energy of the apex. Alpha is alpha1,α2,α3,α4The weight coefficient is adjusted to be 0.1, 0.3, 0.3 and 0.3 through repeated experiments.
In the formula, the Q function calculates the likelihood distribution of the vertexes, the V function calculates the spatial correlation of the vertexes in the neighborhood range, the potential energy of the connected vertexes is calculated in an accumulated mode, if the marks of the adjacent vertexes are different, the V is endowed with a penalty coefficient, the C represents a vertex set, and the f is the solved optimization mark.
In step 5, the specific steps of the tower precise positioning process based on the context relationship are as follows:
step 5.1, converting the linear segment extracted in the step 4 into a binary image;
and 5.2, detecting the pole tower by calculating the maximum direction change and the direction parallelism. Wherein the maximum directional change and the directional parallelism are defined as follows:
maximum direction change (MOV) is the maximum change in direction of all lines in a pixel. For the tower is approximately 90 degrees and the power line is 0 degrees.
The direction parallelism (OP) is the average difference between the direction of the line in the current pixel and the direction of the line in the neighboring pixels.
In step 6, the specific steps of the power line point cloud extraction process are as follows:
step 6.1, setting a distance threshold value between the power line point cloud and the power line segment;
and 6.2, sequentially searching the electric power linear sections, and marking all point clouds meeting the distance condition as power line points.
Claims (4)
1. A method for detecting a laser radar point cloud data power line by linear voxel segmentation is characterized by comprising the following steps:
a ground object point cloud extraction step, wherein ground point cloud is separated from three-dimensional point cloud data acquired by a flight platform, so that ground object point cloud data are obtained;
a linear voxel segmentation step, wherein a linear segment map model is constructed based on linear voxel segmentation;
a candidate line segment extraction step, namely extracting a power line candidate linear segment by adopting a Markov random field model;
a power line point cloud detection step, namely positioning a power line linear section based on the context relationship between a tower and a power line, and extracting power line point cloud through the distance threshold value of the power line point cloud and the power line section;
wherein the linear voxel segmentation step specifically comprises:
step 3.1, carrying out voxelization processing on the ground object point cloud data by using an octree structure, wherein the specific operation comprises the following steps: 1) counting the maximum and minimum values of the point cloud data on the X, Y and Z axes to obtain a three-dimensional bounding box of the point set; 2) the whole three-dimensional bounding box is regarded as an initial three-dimensional grid node, the initial three-dimensional grid node is divided into regular three-dimensional grids, and points in each cube are defined as voxels; if more than three points are contained in the grid node, the grid node is defined as full, otherwise, the grid node is defined as empty; deleting the octree nodes which are empty, and subdividing the non-empty nodes until the termination standard is reached:
R≤Rvoxel||pnum≤minp (1)
wherein R is the voxel volume and is expressed by the side length of a cube, RvoxelIs the set voxel resolution; p is a radical ofnumIndicating the voxel size, i.e. the number of points in the voxel, minpRepresenting the minimum voxel size, and correlating to the point cloud density;
step 3.2, evaluating the linear possibility of each voxel by adopting principal component analysis; firstly, constructing a covariance matrix for point clouds in each voxel, and calculating characteristic values lambda 1, lambda 2 and lambda 3 through PCA; since the linear structure points are linear in space, the eigenvalues obtained by PCA have λ 1>>λ2>λ 3, points on the surface have planar characteristics in space, the characteristic values λ 1, λ 2 have small difference, and the scattering points are distributed and dispersed in space, so that the values of λ 1, λ 2, λ 3 are close, and the method utilizes the principle that
Identifying a linear structure point cloud within a voxel; determining the voxels with linear structures by calculating the proportion of linear structure points in each voxel, if the point clouds in the voxels are all linear structure points, the voxels are linear voxels, if nonlinear structure points exist in the voxels, the linear voxels are compared with adjacent voxels, and if the adjacent voxels are linear voxels and have similar directions with the linear voxels, linear voxel labels are given to the adjacent voxels;
step 3.3, converting points in the linear voxels into linear segments by 3D line fitting based on random sample consensus (RANSAC), and generating a graph model G ═ V, E by combining the linear segments (i.e. vertex V) and the connection between them (i.e. edge E);
the candidate line segment extracting step specifically comprises the following steps:
step 4.1, calculating the height Δ H and the maximum value H from the ground of the linear sectionmaxThe power line is overhead, and the bottom of the tower is in contact with the ground;
wherein H represents the elevation of the linear section, HGIs the ground height;
step 4.2, calculating the slope k of the line, wherein the power line is horizontal, and the rest parts of the tower except the horizontal cross arm are vertical;
where N represents the number of point clouds in the linear voxel, κxi、κyiRespectively representing the fitted curve slopes of the ith point cloud in the linear voxel in XZ and YZ projections;
step 4.3, calculating the point ratio mu of the inner cylinder and the outer cylinder, establishing an inner cylinder and an outer cylinder by taking the line segment as a circular axis, taking the inner radius as 0.15m and the outer radius as 1m, and calculating the ratio of the number of point clouds in the inner cylinder and the outer cylinder; almost no point cloud exists between the inner cylinder and the outer cylinder of the power line;
wherein M represents the number of point clouds in the inner cylinder and M represents the number of point clouds in the outer cylinder;
step 4.4, calculating the parallelism degree sigma of the lines, wherein the single power line is parallel to other power lines, but the towers are randomly distributed(ii) a Suppose the direction vector of the linear segment p is ξpThe size of the search window is set to 20m, the number of linear segments contained in the search window is N, the number of linear segments having the same direction as the linear segment p is m, and the direction vector of the linear segment q is ξqIf 1- | ξq·ξpIf | is less than 0.1, the linear segments q and p are considered to have the same direction, and the linear parallelism σ of the linear segment p ispIs shown as
Step 4.5, aiming at the linear segment graph model constructed in the step 3.3, describing the uniqueness and the context attribute by adopting a Markov random field, and constructing an energy function by utilizing the height of three data items from the ground, the slope of a line, the ratio of an inner cylindrical point to an outer cylindrical point and the parallelism of a context item line; in a graph model G ═ (V, E) composed of vertices V (linear segments) and edges E (connected edges of linear segments), only locally connected edges remain in the graph, i.e. if the distance between a pair of vertices connecting the edges is greater than a specified threshold TdIf 3m, the edge is discarded; according to the state of the current vertex or the adjacent vertex, namely a power line or a non-power line, each energy item has a different energy function, and the potential energy of the vertex is obtained by calculation of a formula (6); searching a power line candidate linear segment through global minimization of an energy function, and iteratively switching the linear segment with the highest energy until global energy is converged; setting the energy function form as shown in formula (7);
wherein U is the potential energy of the vertex; alpha is alpha1,α2,α3,α4The weight coefficient is 0.1, 0.3, 0.3, 0.3;
in the formula, the Q function calculates the likelihood distribution of the vertex, the V function calculates the spatial correlation of the vertex in the neighborhood range, if the labels of the adjacent vertices are different, the V is endowed with a penalty coefficient, and f is the optimized label.
2. The method for detecting the power line of the linear voxel partitioned lidar point cloud data according to claim 1, wherein in the step of extracting the point cloud of the ground object, the step of obtaining the point cloud of the ground object comprises:
a triangular mesh construction sub-step, namely constructing a mesh index for the point cloud data, and taking the lowest point of each block in the mesh as an initial ground seed point; constructing an initial triangulation network TIN based on the initial ground points;
and in the point cloud iterative traversal substep, iteratively traversing the residual point cloud, adding the residual point cloud which meets preset conditions relative to the ground height and the terrain angle into the triangular network, and updating the TIN model of the triangular network.
3. The method for detecting the power line of the linear voxel partitioned laser radar point cloud data according to claim 1, wherein the step of detecting the power line point cloud comprises:
a power line model construction sub-step, namely positioning the position of a tower based on the context of a linear section; constructing a connecting edge set of a tower, namely a power line model;
and a power line extraction sub-step, namely classifying the ground object point cloud data, and extracting the power line point cloud by setting a distance threshold between the point cloud and the power linear section.
4. The method for detecting the power line of the linear voxel partitioned laser radar point cloud data according to claim 1, wherein in the step of detecting the power line point cloud, the specific steps of the process of accurately positioning the tower based on the context relationship are as follows:
step 5.1, converting the linear segment extracted in the step 4 into a binary image;
step 5.2, detecting the pole tower by calculating the maximum direction change and the direction parallelism; wherein the maximum directional change and the directional parallelism are defined as follows:
maximum direction change (MOV) which is the maximum change in direction of all lines in one pixel; about 90 degrees for the tower and 0 degrees for the power line;
a direction parallelism (OP) which is the average difference between the direction of the line in the current pixel and the direction of the line in the neighboring pixel;
the specific steps of the power line point cloud extraction process are as follows:
step 6.1, setting a distance threshold value between the power line point cloud and the power line segment;
and 6.2, sequentially searching the electric power linear sections, and marking all point clouds meeting the distance condition as power line points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911105764.6A CN110794413B (en) | 2019-11-13 | 2019-11-13 | Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911105764.6A CN110794413B (en) | 2019-11-13 | 2019-11-13 | Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110794413A CN110794413A (en) | 2020-02-14 |
| CN110794413B true CN110794413B (en) | 2021-11-16 |
Family
ID=69444272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911105764.6A Active CN110794413B (en) | 2019-11-13 | 2019-11-13 | Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110794413B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111895907B (en) * | 2020-06-18 | 2023-02-03 | 南方电网数字电网研究院有限公司 | Electricity tower point cloud extraction method, system and equipment |
| CN112365570A (en) * | 2020-11-13 | 2021-02-12 | 广东电网有限责任公司 | Method, device and equipment for quickly displaying point cloud of power line channel |
| CN114563796B (en) * | 2022-03-03 | 2022-11-04 | 北京华宜信科技有限公司 | Power line identification method and device based on unmanned aerial vehicle laser radar detection data |
| CN114972672B (en) * | 2022-07-29 | 2023-05-09 | 安徽继远软件有限公司 | Method, device, equipment and storage medium for constructing live-action three-dimensional model of power transmission line |
| CN115508844B (en) * | 2022-11-23 | 2023-03-21 | 江苏新宁供应链管理有限公司 | Intelligent detection method for deviation of logistics conveyor based on laser radar |
| CN117576144B (en) * | 2024-01-15 | 2024-03-29 | 湖北工业大学 | Laser point cloud power line extraction method and device and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104969260A (en) * | 2012-09-27 | 2015-10-07 | 西门子制品生命周期管理软件股份有限公司 | Multi-bone segmentation for 3D computed tomography |
| CN107767389A (en) * | 2017-10-27 | 2018-03-06 | 云南电网有限责任公司电力科学研究院 | A kind of power transmission line extraction method based on airborne laser point cloud data |
| CN108765446A (en) * | 2018-05-30 | 2018-11-06 | 湖北大学 | Power line point cloud segmentation method and system based on random field and random forest |
| CN109147030A (en) * | 2018-07-05 | 2019-01-04 | 厦门大学 | Indoor and outdoor scene based on line feature combines modeling method |
| CN109964278A (en) * | 2017-03-30 | 2019-07-02 | 艾腾怀斯股份有限公司 | Pass through the system and method for the error in evaluated in parallel classifier the first classifier of output calibration |
-
2019
- 2019-11-13 CN CN201911105764.6A patent/CN110794413B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104969260A (en) * | 2012-09-27 | 2015-10-07 | 西门子制品生命周期管理软件股份有限公司 | Multi-bone segmentation for 3D computed tomography |
| CN109964278A (en) * | 2017-03-30 | 2019-07-02 | 艾腾怀斯股份有限公司 | Pass through the system and method for the error in evaluated in parallel classifier the first classifier of output calibration |
| CN107767389A (en) * | 2017-10-27 | 2018-03-06 | 云南电网有限责任公司电力科学研究院 | A kind of power transmission line extraction method based on airborne laser point cloud data |
| CN108765446A (en) * | 2018-05-30 | 2018-11-06 | 湖北大学 | Power line point cloud segmentation method and system based on random field and random forest |
| CN109147030A (en) * | 2018-07-05 | 2019-01-04 | 厦门大学 | Indoor and outdoor scene based on line feature combines modeling method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110794413A (en) | 2020-02-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110794413B (en) | Method and system for detecting power line of point cloud data of laser radar segmented by linear voxels | |
| CN113066162B (en) | Urban environment rapid modeling method for electromagnetic calculation | |
| CN109241978A (en) | The rapid extracting method of planar chip in ground three-dimensional laser point cloud | |
| CN109059791B (en) | Deformation monitoring method and device for power equipment | |
| Su et al. | Extracting wood point cloud of individual trees based on geometric features | |
| CN110363299B (en) | Spatial case reasoning method for outcrop rock stratum layering | |
| Nardinocchi et al. | Fully automatic point cloud analysis for powerline corridor mapping | |
| CN113554595A (en) | Unmanned aerial vehicle laser radar point cloud tower head deformation detection device and method | |
| Shokri et al. | A robust and efficient method for power lines extraction from mobile LiDAR point clouds | |
| KR101221755B1 (en) | Method for identifying reflectivity cells associated with severe weather | |
| CN108765446B (en) | Power line point cloud segmentation method and system based on random field and random forest | |
| KR20120119749A (en) | Method for tracking reflectivity cells associated with severe weather | |
| Miyazaki et al. | Line-based planar structure extraction from a point cloud with an anisotropic distribution | |
| Sun et al. | Automated segmentation of LiDAR point clouds for building rooftop extraction | |
| Yang et al. | Flood Detection Based on Unmanned Aerial Vehicle System and Deep Learning | |
| CN114004740B (en) | Building wall line extraction method based on unmanned aerial vehicle laser radar point cloud | |
| Rufei et al. | Research on a pavement pothole extraction method based on vehicle-borne continuous laser scanning point cloud | |
| Su et al. | Automatic multi-source data fusion technique of powerline corridor using UAV Lidar | |
| Gaha et al. | A new LiDAR-based approach for poles and distribution lines detection and modelling | |
| Dey et al. | Building Boundary Extraction from LiDAR Point Cloud Data | |
| KR101737889B1 (en) | filtering and extraction of feature boundary method from terrestrial lidar data using data mining techniques and device thereof | |
| Chen et al. | Intelligent Interpretation of the Geometric Properties of Rock Mass Discontinuities Based on an Unmanned Aerial Vehicle | |
| Hujebri et al. | Automatic building extraction from lidar point cloud data in the fusion of orthoimage | |
| Husain et al. | An automated method for power line points detection from terrestrial LiDAR data | |
| Shokri et al. | POINTNET++ Transfer Learning for Tree Extraction from Mobile LIDAR Point Clouds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |